Keywords

1 Introduction

Self-sensoring prescriptive applications (SSPA’s), often referred to as “wearables,” have a strong presence in healthcare as a means to monitor and improve health, modify behavior, and reduce medical costs. However, the commercial sector is quickly adopting SSPA’s to monitor and/or modify consumer behaviors as well [1,2,3]. Interestingly, the direct impact biosensor data have on user decision making, attitude formation, and behavior has not been well researched. Social scientists and a number of regulatory bodies have begun to tackle questions about how to manage the reliability and value of these devices [8,9,10,11], however even if SSPA were designed with benevolence and transparency, omnipresent and robust monitoring of the physiological self may still have numerous unintended consequences. One of these is the activation of certain beliefs related to free will and determinism which have been shown to have numerous negative impacts on individual behavior.

Definitions of free will vary, but generally tend to refer to an individual’s belief in their ability to make deliberate choices and the belief that they are responsible for those choices [12, 13]. The following section highlights experimental research as evidence of the possible social impacts of altering individual beliefs through priming and draws attention to how SSPA’s may contribute to altering these beliefs.

Early research related to BFW and deterministic world views demonstrated that attributing learning outcomes to innate qualities of intellect rather than learned behaviors (such as hard work) had a negative impact on personal effort and motivation [5]. However, [6] examined whether the effects of priming such deterministicFootnote 1 beliefs could lead to overt negative moral behaviors. In their study, participants were asked to read a series of statements that either supported a belief in free will, refuted such a belief, or were neutral in nature. Participants then completed a set of problems in reading, math, logic, and reasoning and were told they would receive $1 per correct answer. In some conditions of the experiment participants were presented with an opportunity to cheat by grading their own answers. Results showed that participants who read the deterministic statements and were given an opportunity to cheat took home more money than all other participants.

While research has demonstrated that BFW can influence behavior and perception, to understand how this is related to SSPA’s it is important to understand what environmental cues may reduce or enhance this belief. The studies mentioned above employed written prompts, but other researchers have demonstrated that less explicit cues play a role as well. For example, [15] found that BFW is related in a number of ways to an individual’s perceived ability to make choices. They found that participants who had been asked to recall past choices during a specific time period had stronger beliefs in free will relative to participants asked only to recall specific actions they had taken during a similar time period. In the same study they found that asking participants to make simple choices (in this case, choosing between different pen types) also increased BFW relative to participants who were asked instead to perform a series of simple actions.

Consumer SSPA’s monitor and process a user’s sensor data and implicitly or explicitly direct their actions based on this data. Research suggests that how these directives are presented to the users could impact that user’s BFW. Directives presented as a choice between two or more actions would be less likely to reduce the user’s BFW than directives presented as a single command. For example, an SSPA that informs a user his potassium is low could suggest the user take one of three actions (e.g. eat a banana, drink a glass of fat free milk, or take a supplement) rather than just one.

Even more relevant to the discussion of free will and SSPA is research conducted by [4] regarding physical states of the body and belief in free will. The perception of conscious control over one’s bodily actions could be considered a form of “evidence” to strengthen belief in free will [14], but what happens when control of the body’s actions seems difficult or impossible? [4] first compared the strengths of belief in free will between individuals with medical conditions that cause physical symptoms beyond conscious control (epilepsy and panic disorders) and individuals who did not have these medical conditions. They found that participants who had epilepsy and participants who had a panic disorder had weaker beliefs in free will than participants who had neither condition. In a follow up study they found that more temporary states of the body can also affect BFW. Participants were first asked about their BFW and were subsequently asked about the intensity of some of their physical needs at that moment, including urination, sexual desire, fatigue, thirst, and hunger. They found that participants who had reported more intense needs for urination, sexual desire, or fatigue had expressed weaker beliefs in free will. For hunger, they found that this was also the case for individuals who were not currently dietingFootnote 2. [4] interpret their broader findings as evidence that physical states can influence BFW. They also extrapolated that the less control a person has over those physical states, the weaker their BFW.

[4] does not specify whether they believe physical states impact BFW because the physical sensations create an unconscious awareness of those states (i.e. the stronger the sensation the weaker the belief) or whether reminding a person about their physical state in the past may have a similar effect even in the absence of sensations. In their second study, [4] asked participants about physical states that are typically associated with sensations that the participants could have been experiencing at the time (e.g. hunger pains, fullness of bladder). However, in their first study it is not known whether participants who identified as having a current or past diagnosis of epilepsy or a panic disorders had been experiencing physical sensations at the time they completed the online study or if, instead, the effect was a result of those participants having been reminded about these physical states by being asked to identify as having that particular diagnosis in order to participate. Due to the disruptive nature of seizure disorders and panic disorders, intuitively it seems reasonable to believe that most participants would not have been experiencing major symptoms at the time they completed the study. What this suggests is that reminding participants about physical states that are beyond a person’s control may also weaken BFW. This is particularly relevant to the discussion of the social impacts of SSPA’s, because they are specifically designed to unmask the hidden nature of our internal states. Being reminded (whether through physical sensations or environmental cues) that our free will must sometimes be trumped by our physical needs is an integral part of the human experience and, since most people maintain a belief in free will [12, 13], the weakening effects are likely transitory as these reminders come and go. But ubiquitous SSPA’s offer an unprecedented opportunity to remind users of the countless physiological states of the body that change without willful intent. In this light, SSPA’s may serve to continually depress belief in free will, even those that are genuinely meant to improve general wellness. To investigate this potential risk, this study tests the hypothesis that participants using a wearable activity tracker to monitor heart rate for a short period of time will have reduced BFW relative to participants who explore the menu’s of the device, and to participants in the control condition.

2 Method

Sixty-nine Missouri University of Science and Technology students participated in the study. The research was framed to potential volunteers as a short usability study for a wearable activity tracker and modest compensation was offered. The final sample size included 69 participants with a mean age of 20.49, mean of 2.5 years of secondary education, 67% were male, 81% identified as white, 7% identified as Black/African American, 7% identified as Hispanic, and 5% identified as Other.

The wearable technology used for the study was the Garmin vívoactive® HR which tracks a number of activity types, including step count and heart rate [18]. To measure Belief in Free Will (BFW), participants completed the twenty-seven item Free Will and Determinism Plus scale (FAD+; [17]) that measure layperson’s beliefs not only in free will but also related constructs of scientific determinism, fatalistic determinism, and unpredictability. The items (e.g., “People are always at fault for their bad behavior.”) were rated using a 5 point Likert scale from 1 (strongly disagree) to 5 (strongly agree).

The study used a between-subjects design with two groups plus a control condition where the independent variable was the task performed with the wearable device (Task) and the primary dependent measure was BFW.

Participants were randomly assigned to one of three Task conditions: Heart Rate, Usability, and Control. In the Heart Rate condition, participants were asked to don the wearable device on their wrist for about 30 s and then asked to read their heart rate as reported by the watch. They were then asked to make a prediction about what their heart rate might be if they were to walk at a moderate pace down the corridor of the building in which the lab was located. Finally, participants in this condition walked the corridor while wearing the device and were asked to retake their heart rate when finished. In the Usability task, participants were asked to sit and complete a short list of activities with the watch. The activities (e.g. identify what types of exercises you can track with the device, identify what information you can track about the body) were designed to ensure the user viewed the main functions of the device, without actually taking their heart rate, and spent the same amount of time interacting with the device as participants in the Heart Rate condition. In both the Heart Rate condition and the Usability condition, after the task was completed, participants were asked to complete a survey which was described to the them as containing questions about the watch and questions completely unrelated to the watch.

The first set of questions presented was the FAD + scale [17]. This was followed by a series of questions about design aspects of the device (e.g. look and feel, brightness of screen, ease of use, etc.), but these were primarily meant as cover to confirm the participants’ beliefs that the study was about usability of an activity tracker. The survey also asked participants if they currently own a wearable activity tracker, whether they have a chronic health condition that requires daily management, to what degree they currently felt hungry, tired, in pain, in control of thoughts and feelings, and the need to use the bathroom. Demographic questions were included at the end of the survey. Participants in the Heart Rate condition were also asked how accurately they were able to predict what their heart rate would be after walking the corridor. In the Control condition participants completed the same task and survey as those in the Usability condition, however they completed the FAD + scale [17] prior to interacting with the device. Upon completion participants were compensated and debriefed.

3 Results and Analysis

A factor analysis using maximum likelihood extraction, promax rotation, and four fixed loadings indicated that of the 7-item subscale measuring layperson’s Belief in Free Will (BFW) (included in the 27-item FAD + scale [17]) only 6 items loaded (α = .76). These 6 items were converted to factor scores for each participant and used in further analysis as a measure of BFW, however means scores of these 6 items are included for readability.

An Analysis of Variance (ANOVA) on BFW indicated significant variation across the three conditions, F(2, 67) = 7.73, p = .001, ηp = .197. Since all contrast were of interest and sample sizes were nearly equal across all conditions, Scheffé post hoc criterion for significance were conducted. Participants in the Heart Rate (M = 3.49, SD = .68) condition had significantly lower scores on the BFW scale than the Control condition (M = 3.79, SD = .63, p = .001) but not the Usability condition (M = 4.09, SD = .52) and BFW scores did not significantly differ between the Usability condition and the Control condition.

To better understand these findings a number of other analyses were conducted. First, to determine if reduced BFW in the Heart Rate condition could be explained by how accurate participants perceived themselves to be about predicting their heart rate, a correlational analysis was conducted using the BFW factor scores and Perceived-Accuracy rating (measured using a Likert scale from 1(completely inaccurate) to 5 (completely accurate)), however there was no significant correlation between the two variables (p = .731).

To address the possibility that physical states of the body at the time participants completed the study could explain any changes in BFW [4], an ANOVA was conducted on the BFW factor scores for the five questions asking participants to rate the degree to which they currently felt hungry, the need to use the bathroom, pain, tiredness, and in control of their thoughts and feelings. Only the question about feeling in control of thoughts and feelings varied by condition, F(4, 65) = 3.96, p = .006. However, when an ANOVA on BFW and the Task variable was conducted using this new variable as a covariate, it did not change the significance of effect of Task variable (p = .002).

Other potential covariate including gender, perceived socio-economic status, age, and education were found to have no impact on BFW, nor did current ownership of a wearable activity monitor, or perceptions about the perceived usefulness of wearable activity monitors (all p’s > .05). The presence of a chronic health condition requiring daily management did not have an impact on BFW, however only 17% of respondents responded yes to this question so further investigation with a larger sample size is needed.

3.1 Discussion and Conclusion

These preliminary results suggest that monitoring heart rate with a wearable activity tracker could, at least in the short term, lower BFW. However, it is worth noting that participants in the Usability condition were not told to complete their list of activities in a particular order so they, in fact had a slightly higher degree of choice than participants in the Heart Rate condition. [15] found that allowing participants to make simple choices (such as choosing between different pen types) increased BFW relative to participants who were asked instead to perform a series of simple actions. It is possible that participants asked to complete the Heart Rate task merely felt less choice than participants asked to complete the Usability task. This could explain why BFW scores were lower in the Heart Rate condition than the Usability condition, but it would not necessarily explain why the Usability condition did not have lower BFW scores than the Control condition.

Cutting edge wearables aim to employ sensor data to monitor states of the body that users may have even less control over than heart rate, particularly in the short term, such as vitamin deficiencies and cholesterol. In these cases, an ongoing state of awareness of factors beyond one’s control may act to reduce BFW which could lead to negative affect and a decrease in prosocial behaviors. Planned future research will examine the duration of the effect that heart rate monitoring has on reducing BFW scores and whether this reduction also leads to negative consequences found in past research. Finally, examination of how the design of SSPA user interface impacts this potential risk factors should be explored.

For SSPA’s (and too many other forms of technological innovation) the approach to considering these broader social implications is often to bring products to market, see what sticks around, and evaluate outcomes post-adoption. It is not immediately clear whose should be held accountability for these social implications. Although engineers are centrally located in reflecting on and responding to the ethical implications of SSPA design and deployment, it is critical that regulatory bodies, such as the US Food and Drug Administration and the US Federal Trade Commission, place a higher emphasis on the risks of SSPA, one that goes far beyond the current focus on privacy, data security, and to some degree, sensor reliability.

The economic success of consumer SSPA’s relies on convincing users that their physiological states should not only be monitored, but also controlled. SSPA’s that promise this affordance will no doubt be appealing to a consumer base driven to evaluate, understand, and compare the “self” relative to others and socially constructed notions of normality. It is this affordance and this drive that could lead to unprecedented ubiquity of SSPAs, making it that much more critical to get ahead of potential negative impacts of such technologies before technological inertia takes hold.

On a final note, it will be tiresome to some and critical to others to point out that the research and arguments presented here, though critical of SSPAs, do not, in fact, constitute a luddite call to ban SSPA’s. Rather, it constitutes a call to acknowledge that there may be social behavioral risk associated with SSPA design that are not considered in existing ethical analysis and regulatory processes. And to underscore that the promises of user empowerment and personalized wellness stemming from advocates of self-quantification rest not merely in the expansions of the variety and details of self-sensor data made available to users, but in the design and deployment of the SSPA’s that use these data to influence user behavior and define the self.